Similarity searching of sub-structured databases

Abstract

A grouped data base 10 includes a plurality of sub-databases 10A, in which data have been stored, for respective ones of data sets having specific classification data ("male", "female", "annual income of 20,000,000 yen or more"). The grouped data base 10 is searched in accordance with a given search condition. Given search conditions are stored successively in a search-result data base 11. The sub-data bases 10A contained in the grouped data base 10 are reorganized so as to be in line with those given search conditions that have a high frequency of occurrence.

What is claimed is: 
     
       1. A database search method comprising the steps of: searching a database, in which there is stored a data set containing an identification code and including classification data on an item-by-item basis, by giving a search condition which designates the classification of the data set and using a predetermined search method decided by the given search condition and the structure of the database;   successively storing, whenever a search is conducted, the search condition, search method and time required for a search when a search is conducted;   calculating degrees of similarity between a given search condition and stored search conditions when a search condition has been given;   reading out a search method used when search time, required when a search was conducted under a search condition having a high calculated degree of similarity, is short; and   searching said database by the search method read out and outputting said identification code of the data set having classification data conforming to the search condition.   
     
     
       2. A database search method for a case where there are one or a plurality of sub-databases, created based upon an original database comprising a collection of data sets containing identification codes and including classification data on an item-by-item basis, comprising a collection of data sets having a specific classification in common, wherein the sub-databases are reorganized, comprising the steps of: storing designated classifications contained in a search condition, search methods and times required for the searches whenever the search condition, which designates one or a plurality of classifications to be searched, is given;   calculating degrees of similarity between designated classifications having a high frequency of occurrence among the stored designated classifications and a specific classification common to the sub-databases;   in a case where there is a designated classification among the designated classifications having a high frequency of occurrence that exhibits a low degree of similarity with regard to any specific classification, creating a sub-database comprising a collection of data sets having this designated classification in common;   storing a specific classification common to sub-databases whenever a sub-database is created;   calculating designated-classification degrees of similarity between a designated classification contained in a given search condition and designated classifications that have been stored;   calculating specific-classification degrees of similarity between a specific classification of a sub-database and specific classifications that have been stored;   conducting a search of said sub-database by a search method used when the designated-classification degree of similarity and the specific-classification degree of similarity are high and the time required when a search was conducted at such time is short, and outputting said identification number of the data set having classification data conforming to the search condition.   
     
     
       3. A database search apparatus comprising: database searching means for searching a database, in which there is stored a data set containing an identification code and including classification data on an item-by-item basis, by giving a search condition which designates the classification of said data set and using a predetermined search method decided by the given search condition and the structure of said database;   memory means for successively storing, whenever a search is conducted, the search condition, search method and time required for a search when a search is conducted by said database search means;   similarity calculating means for calculating degrees of similarity between a given search condition and search conditions, which have been stored in said memory means, when a search condition has been given;   search-method readout means for reading out a search method used when search time, required when a search was conducted under a search condition having a high degree of similarity calculated by said similarity calculating means, is short; and   identification-code output means for searching said database by the search method read out by said search-method readout means and outputting said identification code of the database having classification data conforming to the search condition.   
     
     
       4. A database search apparatus for searching one or a plurality of sub-databases, created based upon an original database comprising a collection of data sets containing identification codes and including classification data on an item-by-item basis, comprising a collection of data sets having a specific classification in common, comprising: an input unit for accepting a search condition designating one or a plurality of classifications to be searched;   first memory means for storing designated classifications contained in a search condition, search methods and times required for the searches whenever the search condition is accepted by said input unit;   first similarity calculating means for calculating degrees of similarity between designated classifications having a high frequency of occurrence among the designated classifications stored in said first memory means and a specific classification common to the sub-databases;   determination means for determining whether the designated classifications having a high frequency of occurrence include a designated classification that exhibits a low degree of similarity with regard to any specific classification in the degrees of similarity calculated by said first similarity calculating means;   sub-database creating means which, when it has been determined by said determination means that the designated classifications having a high frequency of occurrence include a designated classification exhibiting a low degree of similarity with regard to any specific classification, is for creating a sub-database comprising a collection of data sets having this designated classification in common;   second memory means for storing a specific classification common to sub-databases whenever a sub-database is created by said sub-database creating means;   second similarity calculating means for calculating degrees of similarity between a designated classification contained in a search condition entered by said input unit and designated classifications that have been stored in said first memory means;   third similarity calculating means for calculating degrees of similarity between a specific classification of a sub-database and specific classifications that have been stored in said second memory means; and   identification-number output means for conducting a search of said sub-database by a search method used when a designated-classification degree of similarity calculated by said second similarity calculating means and a specific-classification degree of similarity calculated by said third similarity calculating means are high and the time required when a search was conducted at such time is short, and outputting said identification number of the data set having classification data conforming to the search condition.